I built an open-source tool that lets you point an AI agent at your local project folder.
It reads your code, runs tests, fixes bugs, and writes new features — all in a disposable
sandbox. You review a diff and approve changes with one click.
Unlike ChatGPT/Claude, it can actually install packages, run databases, and test your real
code. Unlike OpenClaw, everything runs in isolation — the agent can't touch your system.
Any model works: Claude, GPT, DeepSeek, or fully local with Ollama.
MIT licensed, runs on your Mac, your keys never leave your machine.
Unlike ChatGPT/Claude, it can actually install packages, run databases, and test your real code. Unlike OpenClaw, everything runs in isolation — the agent can't touch your system.
Any model works: Claude, GPT, DeepSeek, or fully local with Ollama.
MIT licensed, runs on your Mac, your keys never leave your machine.
reply